Channel estimation for broadband mobile radio systems is generally a challenge due to a large time variance and frequency selectivity of the radio channels in case of fast moving user equipments (UEs). Due to its importance, feedback overhead reduction has been investigated for a long time and has resulted in quite a number of proposals. The most prominent for OFDM(A) systems are:                a) use interpolation in time and frequency leads to a reduced number of CSI estimates, which have to be fed back.        b) codebook based precoding is a technique where instead of the full CSI information only a ‘parameterized’ feedback is being used, which maximizes overall system performance.        c) apply tracking algorithms: there are quite advanced tracking algorithms, especially in combination with codebook based precoding, which have reduced feedback overhead to an level, where COOPA techniques are at least possible.        d) application of subspace based feedback reduction techniques try to adapt feedback to the subspace dimension of the radio channel. The feedback is than only fed back regularly within the subspace dimensions, which will be typically quite smaller than the full dimension.        e) Prediction techniques try to estimate future radio channels based on the past development of the radio channels.        
All these techniques exploit correlations of the radio channel in frequency and time domain. As long as this correlation is high, there will be also a significant reduction in feedback overhead. Difficult to handle are situations with high time as well as frequency variance.
In view of the above-described situation, there exists a need for an improved technique that enables to efficiently determine and feed back channel state information, while substantially avoiding or at least reducing one or more of the above-identified problems.